Two years ago a Spring upgrade took a Navan senior staff engineer over a month and a bad merge locked master for a day. For the Spring Boot 4 migration he planned the upgrade once in Cursor, then delegated each component to Cursor Cloud Agents running on Navan's self-hosted machines, where agents use internal tooling and MCPs for Snowflake and Slack and spin up browsers to verify their own changes. Eight applications and 15 libraries moved at the same time while he monitored the agents, and the migration finished in about a week without freezing development. All generated code goes through the usual human and AI review, testing and staging. About 200 Navan staff now actively use Cloud Agents (reported by Cursor, citing Navan; vendor-published, single migration)
Framework migrations were done serially by one engineer over a month, and incomplete pushes and reverts could lock the master branch and disrupt everyone else's work (reported by Cursor)
An engineer plans the Spring Boot 4 upgrade in Cursor
Each application or library is delegated to a Cursor Cloud Agent running on Navan's self-hosted machines
Agents use internal tooling and MCPs (Snowflake, Slack) and spin up browsers to verify their own changes
The engineer monitors multiple agents in parallel and corrects course
Changes go through human and AI review, testing and staging before production
Human-in-the-loop: The engineer plans the migration and monitors agents; all code passes the same human and AI review, testing and staging process, with complex changes verified in staging.
Spring Boot 4 migration across 8 applications and 15 libraries finished in about a week, versus over a month for a comparable migration two years earlier Reported by Cursor, citing Navan; not a like-for-like comparison · Source (cursor.com)
No codebase freeze during the migration, versus a day-long master-branch lock in the earlier one Reported by Cursor · Source (cursor.com)
Roughly 200 people at Navan actively use Cloud Agents, a 10x increase since April 2026 Reported by Cursor; adoption, not an outcome · Source (cursor.com)
With Cursor, Rahul planned the upgrade in advance, then delegated each component to Cloud Agents running on Navan's self-hosted machines, so 8 applications and 15 libraries moved at the same time instead of one after another.
Once code is generated, it goes through the same human and AI review, testing, and staging process as anything else, and complex changes are verified in staging before production.
Vendor's own product: Microsoft's .NET team reports ten months (May 2025-March 2026) of using GitHub Copilot coding agent (CCA) in dotnet/runtime, the repository behind the .NET runtime, JIT, GC and core libraries. Maintainers explicitly assign bounded tasks; CCA works in a sandbox and opens a PR, and every PR, whether from CCA, staff or the community, also gets automatic Copilot Code Review. Early runs failed because the repo had no copilot-instructions file and the sandbox firewall blocked the NuGet feeds the build needs; the team added instructions and access first. Microsoft reports 878 CCA PRs (14% of all PRs), 535 merged, a 67.9% success rate versus 87.1% for Microsoft engineers' PRs, and 3 of 535 merged CCA PRs reverted (0.6%) versus 0.8% for other PRs (reported by Microsoft; the post says the data is directional, since tasks given to CCA were hand-picked and more bounded)
1Password's engineers hand Codex well-defined user stories, which it breaks into functional specs and near-final prototypes for systems engineers to finish; Codex also reviews PRs before a human, runs acceptance tests in parallel worktrees, and investigates production issues by pulling evidence from incident management, telemetry, source control, paging and feature flags. Security is enforced by keeping secret references (not credentials) in repos and resolving them only when Codex calls an approved internal tool, plus AppSec policies packaged as reusable skills. OpenAI reports a 20.9% measured productivity improvement for the core cohort, a 10.9% cut in median PR cycle time, and investigation of a defect spanning 10+ microservices falling from about two hours to 5-20 minutes. 1Password's 553% ROI and ~$784K annual capacity value are modeled estimates (50 developers, $250K loaded cost, 40% 'directional' Codex attribution, 75% realization) (reported by OpenAI, citing 1Password)
AgentMail's AgentID gives agents email-based identities tied to a human owner so services like Context.dev can allow autonomous agent signups without fraud.
AMD uses GitHub Copilot coding agent to auto-triage 1M+ nightly simulation failures, propose fixes for morning review, and Autofix to resolve ~70% of vulnerability classes automatically, compressing feature cycles from weeks to days across 20,000+ engineers on GitHub Enterprise (reported by GitHub / AMD)